Q2. How do the authors allow for non-linear effects?
In order to allow for non-linear age effects, the authors use a piecewise linear specification of average team age (5-year linear splines).
Q3. How long does the cycle time in the assembly plant vary?
Since one part of the assembly plant (where drivers’ cabs are finished) has very short cycle times (2 – 3 minutes) variation in cycle times in their sample is between 2 and 12 minutes.
Q4. What is the significance of the decomposition of the productivity measure?
A decomposition of their productivity measure into the frequency of errors and error severity shows that the older workers’ competence is their ability to avoid especially severe errors.
Q5. Why do the authors not depend on comparisons across work teams?
Since the authors have nearly 1,000 observations per work team (973 work days spread over four years), the authors do not depend on comparisons across work teams.
Q6. How do the authors correct for the selectivity bias?
The authors correct for this selectivity bias in two ways: first by employing a Heckman-style selectivity-correction model and second by adding worker fixed effects in addition to the work team fixed effects described earlier.
Q7. Why do the authors choose fixed effects for work teams?
Because the authors want to avoid any potential endogeneity of the work team composition with respect to early vs. late shift, the authors choose fixed effects for work teams rather than a fixed effect for each work-station.
Q8. Why is the employer not allowed to monitor workers’ productivity?
Workers cannot affect their wage income through higher or lower effort (i) for contractual reasons and (ii) because their employer is not allowed and not able to monitor workers’ productivity.
Q9. What are some examples of how to measure the contribution of a team to its productivity?
Examples for such potential contributions to a team’s productivity are the instruction of younger workers,4 being relaxed in tense or hectic situations, and contributing positively to the work climate.
Q10. What is the effect of age on productivity?
for workers who grow old in the plant, the productivity enhancing effect of accumulating more experience compensates the adverse “residual” age effect so that the overall age profile is rather flat.
Q11. How do the authors correct for the selection bias in the regression analysis?
Since their observation unit in the regression analyses is a work team while selection into the sample is an individual phenomenon, the authors aggregate individual Mills ratios to team Mills ratios (see Appendix C for details).
Q12. How long does it take to separate age and cohort effects?
While the relatively short observation period of 4 years guarantees the absence of technological change and thus time effects, the statistical basis for separating age and cohort effects is weak.
Q13. What is the effect of population aging on economic growth?
If the impression were true, population aging would have negative effects on overall productivity as the share of older workers is increasing, and would thus directly reduce economic growth.
Q14. What is the coefficient of the interaction term between workload and age?
Its coefficient is negative and insignificant, implying that, if anything, older workers are at least as good as younger workers in dealing with the higher workload generated by a faster running assembly line.
Q15. What is the difference between the age composition of a plant and the productivity of established companies?
As pointed out in the introduction, the age composition of a plant tends to be endogenous to labor productivity since, e.g., fast growing start-ups have freshly hired and thus typically younger staff than established companies.
Q16. Why are errors at different work stations strictly comparable?
To a first degree, errors occurring at different work stations on the assembly line are strictly comparable because every error is given a severity weight that accounts for the costs of fixing that error.